# Kynver

## Connect a coding agent to Kynver chat
Claude Code and Codex can answer your Kynver chat from your own computer. This uses the Kynver runtime, not MCP.

Needs: Node.js 24.11 or newer, Claude Code (`claude`) or Codex (`codex`) signed in, and an approved Kynver workspace. Access is early access, by request: https://kynver.com/agent-os/early-access

Steps:
1. Install: `npm install -g @kynver-app/runtime`
2. Run: `kynver`
3. The first time, it opens a page in your browser. Approve this computer there. No keys to paste.
4. If `claude` is on your PATH and signed in, Kynver finds it by itself. Kynver chat then shows a "Claude Code" tab. A signed-in Codex becomes the engine for website chat the same way.
5. Check what was found: `kynver agents status`
6. Pick the chat folder. The Claude Code tab only works in a folder you allow. Choose it on the Computers page in Kynver, or run: `kynver setup --chat-access folder --chat-folder /path/to/folder`
7. Turn it off: `kynver setup --no-chat-oauth` turns off website chat on this computer. `kynver setup --chat-access off` turns off computer access. `kynver stop` stops Kynver.

Notes:
- macOS and Linux run it directly. On Windows, use WSL, or Kynver Desktop once it is offered in Account > Settings.
- Quit Kynver Desktop before running `kynver`. Only one runtime can work for a workspace on a computer.
- Messages go from the browser to your computer over the connection `kynver` opens. Nothing on your computer is exposed to the internet.

More: https://kynver.com/runtime

## Connect a coding agent to Kynver tools (MCP)
MCP gives your agent your Kynver workspace tools: memory, tasks, plans, contacts and skills. It does not let you chat with your agent from Kynver. For that, use the runtime above.

Create a key under Account > Settings > API keys, then put it in an environment variable. Never paste a real key into a shared file or a chat.

- Claude Code: `claude mcp add --transport http kynver https://www.kynver.com/mcp/kynver --header "Authorization: Bearer $KYNVER_API_KEY"`
- Codex: `codex mcp add kynver --env KYNVER_API_URL=https://www.kynver.com --env KYNVER_API_KEY=<your-api-key> -- npx -y @kynver-app/mcp-agent-os`
- Restart your agent (in Claude Code, run /mcp) so it sees the new tools.

Where work goes:
- Estimating work (estimates, bids, quotes, takeoffs, work orders, change orders, proposals with pricing, and their PDFs) goes in the Estimator, with the estimator_* tools. Add the Estimator server with the same key: `claude mcp add --transport http kynver-estimator https://www.kynver.com/mcp/estimator --header "Authorization: Bearer $KYNVER_API_KEY"`. Its PDF comes from estimator_export_pdf.
- Reports are for projects: status, research, briefings. Kynver refuses an estimate published as a Report.

Agent verification tools (look up and verify AI agents, not your workspace) are a different server: `npx -y @kynver-app/mcp-server`. More: https://kynver.com/mcp

## About
Kynver is the identity layer for AI agents. We help developers and users trust and verify AI agents through a five-pillar Verification Standard.

https://kynver.com

## Verification Standard
The Kynver Verification Standard defines the trust framework for verified AI agents: identity, transparency, behavioral safety, data & privacy, and operational reliability.

https://kynver.com/standard

## Agent Directory
Browse and evaluate AI agents. Filter by category, verification status, and trust score.

https://kynver.com/agents

## Pricing
Developer tiers for agents: Free, Starter, Pro, Business. Verification badge at higher tiers.

https://kynver.com/#pricing

## Blog
Updates, policy, and commentary on AI agent identity and trust.

https://kynver.com/blog